off-topic-tech

[visual tree]

Here are prices btw (used google translate on image so translation might not be perfect):

JT - universal tariff, VT - higher tariff, NT - lower tariff

I am currently on white tariff model

[night girder]

in Belgium, 1kWh is averaged around € 0,28.

[visual tree]

With those prices, everyone would go broke here lol

[night girder]

For this month it could be around € 0.25

[visual tree]

I guess our prices are lower since we started investing in renewable energy production and had energy surplus after many years

If we still had to import energy from Austria, I bet prices would be much higher

[edgy hazel]

Our prices are about the same

[visual tree]

Is it because of shutting down nuclear power plants?

[safe trench]

just eat the raw UraniYUM!™️ waste

[edgy hazel]

Coal lobby really did a big one with that

I remember there was a huge fear campaign about that in 2005-2010

And even with solar getting cheaper and cheaper, it's practically useless because we have basically no infrastructure to transport power across the country

BUT Lufthansa still exists after we bailed them out for the 500th time

[dire igloo]

Didn't the biggest and most effective anti-nuclear fear campaign start with Fukushima

[dire igloo]

Remember: Lufthansa are too big to fail.
Not in the sense that they won't fail because of their size, but because their pockets are too deep to allow them to fail

[edgy hazel]

Oops sry mister goberment... we spent all our monies paying out dividends and now we can't pay our workers : ( pwease give me 6 billion Euros

[languid gulch]

pretty sure the Three Mile Island incident was the most successful fear campaign, as all the other ones started afterwards & the media gobbled it up

[night girder]

New vulnerability discovered in Linux: SLUBStick. Can read/write, escape environments and change privileges.

Paper: www.stefangast.eu/papers/slubstick.pdf

[dire igloo]

eci.ec.europa.eu/045/public/#/screen/home

This initiative calls to require publishers that sell or license videogames to consumers in the European Union (or related features and assets sold for videogames they operate) to leave said videogames in a functional (playable) state.

Specifically, the initiative seeks to prevent the remote disabling of videogames by the publishers, before providing reasonable means to continue functioning of said videogames without the involvement from the side of the publisher.

The initiative does not seek to acquire ownership of said videogames, associated intellectual rights or monetization rights, neither does it expect the publisher to provide resources for the said videogame once they discontinue it while leaving it in a reasonably functional (playable) state.

[visual tree]

Just ditched my old ISP and I'm moving to another since they are running fiber in the streets

I bet salespeople are going to bombard me with calls and make "great offers" so I could stay with them

I have a very simple criteria: the first ISP to connect my home to fiber network gets a contract

[stray badger]

it takes a bit for them to start operating said fiber

its been a month and still no actual internet offered

[visual tree]

My new ISP will finish fiber construction by the end of the year. I will be using vdsl until then

If I survived on copper for 20 years, I can take 4 more months at maximum lol

I still have nightmares of pre-2014 era when I had adsl and speeds under 4 Mbit/s

Or even worse, dial-up era when I was constantly fighting with my sister because she was always on the phone and I couldn't use internet while she was talking lol

[twin dew]

Finally it works, Get the IPs for the possible DNS servers and add those to IPSet for firewall use:
`#!/bin/sh

if serverdns=$( timedatectl show-timesync --property=SystemNTPServers | cut -f2 -d"=" ); then
if serverips=$( resolvectl query -p dns -t A --legend=false ${serverdns} 2>/dev/null); then
serverips=$( printf "${serverips}" | awk '{print $4}' | sed -z 's/\n/, /g' );
nft flush set inet nftables.service ntpservers4;
nft add element inet nftables.service ntpservers4 { ${serverips} };
fi
if serverips=$( resolvectl query -p dns -t AAAA --legend=false ${serverdns} 2>/dev/null); then
serverips=$( printf "${serverips}" | awk '{print $4}' | sed -z 's/\n/, /g' );
nft flush set inet nftables.service ntpservers6;
nft add element inet nftables.service ntpservers6 { ${serverips} };
fi
fi`

Ah, whops, need to move that cut so that errors in the first step also stop the processing.

But just moving that cut to another printf |.

And seems I should be able to replace those awk | sed pairs with just awk...

Well, that was simple... serverips=$( printf "${serverips}" | awk 'ORS=", " {print $4}' );

[twin dew]

And both that and HTTP(S) version for distro updates set up with systemd units.
Both "permanent" ones with stuff binding them to other units, and timer triggered ones.

NTP IPs getting updated every 15 minutes, HTTP ones every hour, and just before automatic updates and upgrades run.

[twin dew]

Ah, I can even do same for DNS, as the local one will only even contact the set one, the iteration/recursion is done by the next DNS server.

Basically I'm locking even outgoing traffic down.

For the linux VMs and machines in DMZ, and the firewall machine itself.

And to complete that, would need to see how to set DNS resolution to only be allowed for whitelisted domains...

DNS queries have been successfully used to exfiltrate data from locked down systems.
query something like <data>.<serial>.yourdomain with blocks of data as payload and log at your authorative name server, then reassemble.

[dire igloo]

you'd want an X3D CPU for SF and Factorio.
7800X3D is the best pick there

also, H9 Flow is hardly ever viable, what about it made you pick that one over other similar cases?

What do you need the cores for?

7900X3D performs worse because it's 6+6 CCD (essentially two 7600X taped together, one of them got extra cache)

Where is 7800X3D has all eight cores in one big cluster, so all of them can quickly talk to each other and access 3D cache

GPU: best you can afford

It's the last part I pick on a build list

[night girder]

3D cache is cache on CPU itself.

[dire igloo]

I stopped recommending H6 Flow because it's a nightmare for compatibility checking.
There are plenty more good fishtank cases on the market, you don't need to pick NZXT

[cosmic imp]

Also maybe hold off on anything past 13th gen on intel

[willow pike]

cache is super fast memory that every CPU has. 3D cache is a technique to stack cache on top of cache, hence X3D, very good for gaming

[dire igloo]

And X3D means it has a buttload of cache - which helps in certain games (like Satisfactory, Factorio and similar)

[night girder]

Yes, but not every CPU has it. I though cache was normally besides the CPU on mobo?

[dire igloo]

The usecase mandates X3D, so not a factor

Depends on the country you're in and your aesthetic preferences

[willow pike]

not since the 90s

fuck i'm old

[dire igloo]

No, cache is integrated into the CPU.
RAM is basically the extension of cache, much slower but also a lot more of it

[willow pike]

but yeah 7800X3D is the meta

[dire igloo]

Do you want a fishtank case?
This means that you have not only a glass side window, but also a glass front window

[night girder]

So, they basically moved the cache on the CPU itself?

[dire igloo]

Way back before the 2000s I guess

Idk, I wasn't around back then

[willow pike]

i was

[cosmic imp]

I just meant in general seeing as their microcode shat the bed on their 13th and 14th gen CPUs

[night girder]

No, that's not what I meant 😄

I google around for the answer. I want to know what the architecture was before 3D, where the cache was position on the CPU.

And where it's positioned now on the CPU.

[willow pike]

next to the CPU cores

[dire igloo]

H9 is basically a generic fishtank case.
I'll make some alternative suggestions and you tell me what you think:
Antec C8
Montech XR
Montech King 95
Phanteks NV5
Magniumgear Neo Qube 2
Lian Li O11 Dynamic EVO
Lian Li O11D Vision

[dire igloo]

Cache has always been on the CPU for the last decades.
What AMD did with 3D cache was that they just stacked cache on top of cache to get more cache into the CPU

[night girder]

You sure?

[willow pike]

vertical stacking like this is the cutting edge, TSMC are very proud of it

[willow pike]

yes

[night girder]

Instead of laying the cache next to the processor, as has been traditionally done, AMD is stacking the cache on top to squeeze more on the chip.

[willow pike]

that's worded for the layman

it has been on the same die as the CPU core/cores for decades

[night girder]

That's confusing AF.

[willow pike]

though, being anal, we could say some cache has always been on the same die, the L1 cache

[night girder]

Sorry, that sentence got me confused. I though literally it changed position on the CPU.

[dire igloo]

Being what?

[willow pike]

do not

[dire igloo]

It always does, core layout is one part of CPU architecture

[willow pike]

every CPU core has an L1 cache, level 1 cache, to hold what it's immediately working on. L2 cache is bigger and slower, L3 cache even bigger and slower but (usually) shared between multiple cores. it's the L3 cache that AMD are stacking

[willow pike]

in ye olden days, 2-core CPUs would share L2 cache and not have an L3 cache

[night girder]

That is on top though ...

[willow pike]

it's on top of cache

64MB on the top chip, 32MB on the bottom chip with the cores

96MB total

the cores and their L1/L2 are the bits on the side without anything on top

[night girder]

So technically, it is on top?

[dire igloo]

Basically: you take a regular CPU and you put a whole new layer of cache on top of it

Sole purpose: MOAAAAR

[willow pike]

the technology isn't AMD invented, it's TSMC SoIC

though i think only AMD use it right now

[night girder]

pretty dope

[willow pike]

cutting edge

[dire igloo]

Makes cooling the CPU a nightmare tho

[rustic panther]

IIRC it's literally on top

[dire igloo]

Which is why X3D chips are notorious for running hot despite not drawing much power

[willow pike]

now see what they use it for on this absolutely yuge machine learning chip

[rustic panther]

Also, I own a 5800x3d and it works really well for satisfactory

[willow pike]

as do i, is very good

[rustic panther]

Yeah, I slightly undervolted mine IIRC so it runs a lot cooler and I don't think I lost much performance

[dire igloo]

I'm holding off on getting 5700X3D, I enjoy my dual _800XT

[willow pike]

@night girder this is an easier to understand diagram

[rustic panther]

The 5000 series from what I've seen also has a tendency to hop right into bed with the thermal limit in general

[night girder]

7800X3D here

[willow pike]

you have the meta CPU

[night girder]

waiting for SF release to test it.

[night girder]

My other CPU wasn't stable. 5800

[rustic panther]

Like if you have AM4 and don't want to shell out for a new motherboard, the 5800x3D is an insanely good CPU still

[willow pike]

RAM tuning can be a bitch on ryzen 5000 in my experience, that may have been the source of it not being stable

[night girder]

nah, was CPU.

[willow pike]

heck, did they replace it

[dire igloo]

5700X3D is usually the better pick

[rustic panther]

That exists now? Fair enough then

[dire igloo]

Only like 10% performance sacrifice but depending on market over $100 savings

[night girder]

We did a full reset of everything and found out the FCLK was messed up. Had to lower it in bios* to get it more stable. But CPU was degrading and became more and more unstable.

[rustic panther]

Haven't been into the hardware game for a while tbh, pretty happy with what I have

Other than laughing/cringing at Intel for fleecing their customers that badly

[night girder]

Should run at 2000MHz, I had to put it around 1733MHz I believe. Was stable for a few months. Then it got worse. Baldur warned me for it, so I bought new CPU for that case.

[dire igloo]

Anyways, back to PC build advice
@cobalt ivy did you look into the cases I sent?

Also, please state the country you're in, the Eurozone is quite big and the markets vary a lot

[willow pike]

baldur is wise

[night girder]

Yeah, he helped me a lot.

[willow pike]

here's a pentium 2 with off-die cache chips

[rustic panther]

Cpu cartridge my beloved

[willow pike]

and that's 512KB cache

[night girder]

I have Lian Li case, love it.

[willow pike]

i hope baldur never does anything wrong because i am not ready for baldurgate

[dire igloo]

Surprised to hear that actually, would've expected you to pick one of the pillarless designs instead.

Anyways, what country and why those over the rest?

[dire igloo]

And that's my friends is why we say model over brand

[night girder]

160mm front fans

[dire igloo]

Och neeeeeeee

[night girder]

and a good case to work in.

[dire igloo]

Na gut, muss sein

[night girder]

Fireworker, you ok?

Oh no, an invasion.

[dire igloo]

Serverregeln besagen dass der Server exklusiv Englisch ist, also am besten schnell zurück wechseln

Aight, so. What makes the O11 series so special against the rest?

[night girder]

https://www.youtube.com/watch?v=6_w0NbB84P0

[dire igloo]

Wha... What?
I didn't understand those words

[night girder]

That's one of the reasons I bought it.

And someone suggested it 😂

[dire igloo]

This didn't clear it up any further.

You mean like the side panel covers the whole side and the PSU is in the back?
Instead of having a PSU compartment in the bottom which the glass doesn't reach down to?

Okay, so why Lian Li O11 over Antec C8 and Montech King 95?

That's just the pcie extension slots, what about them?

I don't get it

Ah, that's what confused me.
You said both Lian Li cases, but the vision is the one you liked most?

Okay yeah

Bottom PSU compartment

So O11 Vision was your case of choice because it had the PSU on the back rather than on the bottom and because it had a top window.
Correct?

Getting too late for me, time for sleep

[safe trench]

why even fear about nuclear power

those accidents were long ago

[languid gulch]

eh, they can still happen, but i'd feel a lot better if we started building LFTR reactors

now that they've officially solved the pipe corrosion issue

[gilded helm]

The Wikipedia list of nuclear accidents is honestly pretty short. And compared to the damage coal and other fuel burning energy has done, nuclear should be a pretty obvious choice.

Of course, you also need to overhaul the regressive nuclear policies of places like the USA too.

[languid gulch]

my biggest local annoyance is that the nuclear plant i'm tied to has already been cleared to add 2 more reactors, and that was decades ago. still haven't chosen to break ground on it

[gilded helm]

Utilizing the existing design? Gen 2 or gen 3?

[languid gulch]

uhhhh no idea

https://en.wikipedia.org/wiki/Palo_Verde_Nuclear_Generating_Station

ah, gen2

NOT RBMK

[gilded helm]

Really the big thing that needs to change is waste reprocessing. Nuclear weapons proliferation is a valid concern and all, but honestly, we're well past that tipping point.

[languid gulch]

i'd like thorium reactors as a much easier fuel to get ahold of since we're already mining a crapton associated with iron & other heavy metals

no need to mine specifically for uranium

[stray badger]

only problem is that they still have not been cleaned up in many places

[languid gulch]

yea, also why i like the idea of using thorium. basically no chance of a meltdown unless you really try, & it's fairly low-density radioactivity compared to uranium

not being able to make nuclear weapons out of the leftovers is also a really nice bonus

it'd let the world know your exact intentions if you chose to build one

[gilded helm]

It's only a concern with new nations obtaining nuclear weapons IMO.

[languid gulch]

was also 100% the point of the first nations building them too

[gilded helm]

Huh, Palo Verde can take MOX

[modest isle]

humans fuck up. there will be more as we build more

but in any case the more compelling argument is that nuclear power to date is the most expensive power generation technology, and idk about you but everyone i know has been feeling the squeeze of inflation latelly so higher electric bills aren't recieved well

if it could pay for itself that would be differnet, but so far nuclear power is a huge money sink and nobody has yet proved otherwise

[safe trench]

or SMR reactors sincethey can be put at more places

[safe trench]

its not like theres one gonna be built right next to my house or in my city

[stray badger]

Bruh there was meltdown in the hills of LA

Still not cleaned

[safe trench]

still

its not like its major

and with newer tech n stuff

its not like they are that wide spread

we would have much better tech when they get wide spread

• its not like if you were to use an smr reactor if it were to melt wouldnt just destroy an entire city

[modest isle]

yeah nuclear accidents are rare, but they leave pieces of land uninhabitable for generations when they do so it's not something to gloss over

[safe trench]

but its also not like its something to make the main reason about nuclear accidents + with newer reactors its not like they are not doing anything to make the event of an nuclear accident much much worse than it was back then in 1986 or 2011 right?

its not like the united states or united kingdom are building an nuclear reactor right beneath the whitehouse or sum right?

[modest isle]

very expensive energy + nightmare accident that costs billions of dollars to remediate the companies don't have when it goes wrong

[safe trench]

im not talking about expensive right now

[modest isle]

yeah there is hope they can make it affordable, but we will see it when they make it happen

[safe trench]

im talking about its not like future reactors wont have meltdowns as terrible as 2011 or 1986 isnt it?

chernobly ruined the scene for nuclear power

[modest isle]

yes they should be able to avoid those situations in future accidents, in theory

still no news on the affordable energy front rho

[safe trench]

true

on the other hand

[modest isle]

if nuclear power were a well proven technology that is a truly a no-brainer to install then then the usa and europe would be increasing nuclear capacity instead of decreasing it

but serious problems remain to be resolved

[safe trench]

it certainly is a lot cheaper and a lot less planet harming to use nuclear power than just build a bajillion factories for masks just to make the air breathable in the far future isnt it tho?

[modest isle]

masks? lol just breahte in the pollution and hope for the best

[safe trench]

my satisfactory world:

[modest isle]

you saw how people reacted to COVID so you know how people will react to shitty air quality..

[safe trench]

pft

imagine having to go to costco when your freind doesnt have any toilet paper and then theres no toilet paper at costco, yeah thats terrorizing

[modest isle]

no need to imagine, i lived it. it was surreal

[safe trench]

ikr

[modest isle]

started to wonder if i was gonna be wiping my ass with leaves

[safe trench]

well thats off topic about what we are talking about

but the ammount of deaths caused by a coal power plant like extraction and delivery coal is way worse than a nuclear power plant

like just imagine if someone begins doing an act of terrorism on the train thats transporting the coal there

its not like the plants the onlything causing emisions

[modest isle]

yes it is true coal power kills far more

[safe trench]

the train and car there needs carbon fuel too doesnt it?

[modest isle]

but i think that is probably a small factor in the caclulations of the relevant authorities when compared to the cost

people die all the time for various reasons but if people cant pay their bills u get voted out of office heh

[safe trench]

heh, thats true

[safe trench]

mrbeast will pay for nuclear power 💯 💯

[modest isle]

lol hes gonna need more suscribers!

[safe trench]

its a shame tho since social media loves to dramatize small accidents

[modest isle]

yeah, but at the same time i think decision makers in places of power tend to see thru these things and put $$$ first, or at least that's my impression

[safe trench]

dont worry, they all do

[modest isle]

we had a train spill tonnes of vinyl chloride monomer in a town here not so long ago and what changed? not much of anything

[safe trench]

its not like my water is now red green and blue color shifting isnt it?

[modest isle]

as long as it doesn't smell like bananas you're good 😉

[safe trench]

if a coal plants turbine were to implode

social media would begin going insane saying "IS THIS THE END OF COAL POWER!?!?!"

[modest isle]

probablt

social media is just a bunch of knee jerk reactions after all

[safe trench]

you know how the whole world reacted to 9/11 didnt you?

[modest isle]

yes but not really sure in what way. i know there was a good deal of celebration in the middle east tho

[safe trench]

DAMM BOI YOU GOT ME LAUHGING

[modest isle]

well we earned that one...

[safe trench]

i swear tho

it wasnt even that big of a deal

its not likesome kind of a380 is gonna crash into the one world trade center again isnt it?

its not like osama bin laden is still alive

[modest isle]

the huge increase in airport security is the important change imo. still lots of people who have good reasons to attack the USA

[safe trench]

airport security got worse

bro, they almost said i was a terrorist for bringing an nintendoswitch to the plane 💀

[modest isle]

the bomb sniffing machines and body searches make i worse somehow?

i doubt it

it's a pain in the ass to go through, yes, but i have little doubt it is more secure

[safe trench]

not that

im talking about tsa tho

[modest isle]

i have actually only flown ONCE since 9/11

and it was annoying

i did sit next to a veteran airman though and he was laughing at me when we hit some turbluence and my voice went all high pitch XD

[safe trench]

lol

but its not like the one world trade center is indestructible

if only some guy figured out how to invent a solid that can explode

[modest isle]

meanwhile the TSA is having an aneurism over my 2 oz. of toothpaste

[safe trench]

so relatable

tsaputs a whole fit over my galaxy note 7 >:(

[modest isle]

lame

[safe trench]

we spent nearly an whole hour talking here

[twin dew]

Cache on MB hasn't been a thing since 90s...

And that P2 era was probably last one with significant amount of cache outside the CPU die, except the L4 eDRAM in the couple of 5th Core models.

[languid gulch]

the scenes in Home Alone of them running thru the airport are wild comedy now

[twin dew]

On P2 the L1 was in-die, L2 was in the CPU module PCB, L3 didn't exist.

Ah, early P3 still had external L2.

Same for first gen AMD Athlon in 1999.

[languid gulch]

i do find it interesting that Gelsinger has been at intel for both of its major fuck ups

[willow pike]

and this is the fucker they said would save them

[vale wigeon]

Anyone has a good keyboard advise? I have a APEX 5 now, with blue switches but those are to clicky and make alot of noise.

Actually checking if i should go with wooting60he

[twin dew]

Browns are the not as clicky blue replacement with still tactile point.

Blues are intentionally also clicky in sound.

But depends on if you actually want that tactile point or not.

[vale wigeon]

Hmm will check thanks!

[dire igloo]

60HE is nice, but linear.
So smooth travel from top to bottom

[night girder]

We've been over this. But thanks for confirming what has been said.

[twin dew]

And I have bad habit of replying to comments before I have read all of the backlog.

[tough owl]

https://www.youtube.com/watch?v=rttc_ioflGo

🤣

[willow pike]

seems it's in eco mode by default

i always turn on eco mode

[tough owl]

/www.tomsguide.com/computing/online-security/29-billion-hit-in-one-of-largest-data-breaches-ever-full-names-addresses-and-ssns-exposed

[twin dew]

I'm somehow sceptical of that 2.9 billion people number...
At least that they had actual data from that amount of people.
When currently world total population is about 8.1 billion.

But to the review, would have been nice to see also raised power limit results.
But someone else probably has those.

And reviews like this are the reason why going back down on power usage is very hard to implement.
And why Intel is stuck with that.

Phoronix with Linux testing:

The Ryzen 7 9700X delivered 1.195x the performance of the Core i5 14600K competition or 1.15x the performance of the prior generation Ryzen 7 7700X. The Ryzen 5 9600X came in at 1.35x the performance of the Core i5 14500 and 1.25x the performance of the Ryzen 5 7600X. Or if still on Zen 3 for comparison, the Ryzen 5 9600X was 1.82x the performance of the Ryzen 5 5600X.

[stray badger]

i also have an apex 5, it is not actually mechanical by the way. it just comes pretty close to feeling like one

the sound does get annoying tho

[twin dew]

But the new things in Zen 5 are things that probably give big efficiency boosts in data center.
Which is the money maker currently AFAIK.
And point in the design was to set stage for next few generations at least.

We will see how the 9950X does against the 7750X in multicore benchmarks.

[stray badger]

Im happy i bought my 7950x3d when i did

[dire igloo]

7750x?

[tough owl]

7950x

I think Baldur's tired

[cyan crescent]

I got the 7900X3D

[twin dew]

Yes, I'm very tired right now.

[charred pewter]

get some rest, let the crazies take over for a bit

[twin dew]

I just woke up way too early today and couldn't get back to sleep.
Going to bed as soon as room temperatures drop.

[willow pike]

get some sleep, or your brain will degrade like an i9

[dire igloo]

gotta love the lack of 65W Zen 4 CPUs in the efficiency chart

[tough owl]

yea

Some people are giving it glowing reviews others aren't

[twin dew]

Stagnated performance for most part at stock on Windows, but significantly better efficiency and lower power.
On linux with optimized binaries there is significant performance uplift, same if you enable PBO there can be 20% improvement on Windows side.

From the stock performance.

So if you only look at the absolute performance, that is +-0 on average.

With often up to 40% less power XD

When the stock 9700X took the performance per watt crown in the GN testing from ECO mode 7950X in multicore test...

www.numberworld.org/blogs/2024_8_7_zen5_avx512_teardown/

Why were Zen5 IPC leaks all over the place?
Leaks about Zen5's IPC improvement have been all over the place. And while AMD has officially stated that the average IPC improvement over Zen4 is 16%, the numbers being averaged behind it are as random as throwing darts after being spun around in circles blindfolded.
Numbers ranged from a low as 5% (mostly Zen 5% memes) to 40% SpecInt, to 2x AVX512. And while it may seem obvious to some, there is a simple reason behind it - there are many different types of workloads and Zen5 improves on them very unevenly. Some things improved a lot while other things gained virtually nothing.
How a benchmark performed depended on where it landed. And it didn't help that Zen5 desktop and Strix have different AVX512 implementations - thus contributing even more variation to benchmark performance.
If we look at pure homogenous CPU workloads with no memory bottleneck, here's what I get from just my own tests (largely taken from my own projects):

[visual tree]

Ah, yes. The great toilet paper war during the pandemic, millions of people were harmed

[dire igloo]

I want whatever drugs the chart guy was on

this sorting is cursed as fuck

[twin dew]

But that flat SSE performance also explains why the Zen4 > Zen5 is static in games.

And that doubled AVX512 performance would make the Zen5 Epycs must-buy for lot of things with bespoke optimized code.

In fact, the AVX512 improvement on Zen5 created a memory bottleneck so large that it became the primary reason why I promoted the BBP mini-program from a tool for verifying Pi records to a formal benchmark. The regular benchmarks wouldn't do Zen5 (and future processors) any justice. At least until someone can figure out how to get DDR5-20000 on AM5...

[sharp oasis]

Yeah that's cursed

[willow pike]

ryzen 7000 was pushed too hard on power compared to prior generations, 9000 is a return to efficiency by default, so it looks like no improvement

though amd should make it much easier to enable PBO and go to higher wattages if people want

one button press in ryzen master rather than digging in the bios

[willow pike]

what the fuck is this

[dire igloo]

do have a 65W vs 65W efficiency comparison for 9000 series?
cuz GN conveniently left out Zen 4 non-X and only kept the 105W TDP CPUs

[twin dew]

Also the Zen 5 big improvements are in AVX512, with some in the lower stuff and even some regressions.
Based on that numberworld entry there might be CPU bug in Zen5 that causes most SIMD instructions to take 2 cycles and not just 1 like previously as example.

Causing havoc on code compiled expecting 1 cycle latency.

And that SSE is static in performance will also downgrade game performance for most floating point code, as lot of games still don't use even AVX.

U8 0.8.2.0 was the first Satisfactory to use any AVX for example.
And no idea how much it uses vs. SSE.

Or if it even still uses AVX, or just SSE4 version that isn't in Phenom 2:s.

But Ryzen 9000, not really for gamers at this time.

[willow pike]

they only just added sse2 to team fortress 2

[safe trench]

yes.

[wanton orchid]

man sse family sets are almost closer to the deprecated phase at this point

AVX512 256bit vectors are the way to go

2010 cpus already had support for avx2

[languid gulch]

i always keep a roll in my car

[wanton orchid]

and there is misconception that avx512 is meant to be used like gpu code
when it's advantages is actually being very low latency mixed cpu calculations and algorithms

and that amplify the thing spotted on avx512 benchmarks

they basically misuse avx512

if you want to do massive data flow calculation you simply offload to gpu or npu or the like

there is no point in doing that in cpu

[stray badger]

is the non x even out?

[dire igloo]

9600X is essentially a 65W TDP with how it acts

so it'd be fair to not match it against 105W TDP 7600X, but a 7600 that has 65W

however, GN left out all of the previous AM5 CPUs with 65W TDP in their power efficiency testing

[stray badger]

I guess, but at that point amd should just market it that way, drop the x

7000 actually was pretty effecient, idk why they were given higher wattage anyway

[cyan crescent]

AMD or intel?
AMD or Nvidia?

[stray badger]

i went amd amd

i simply cannot afford nvidia lmao

intel is... not good atm

[dire igloo]

Intel GPU?

[dire igloo]

12600kf is still bae and 12400F has made am4 builds obsolete at low budgets

[dire igloo]

the answer really depends on budget

[stray badger]

True

For my budget, amd made more sense, but that was only because i had more money than i had expected

[dire igloo]

anything up to 12600KF, Intel
anything beyond that, AMD

GPU wise, Intel for anything up to A770
AMD for the inbetween space until 7900GRE
then whatever bets fits your usecase

[cyan crescent]

Cpu

We all know intel gpus aren’t that good rn

[dire igloo]

idk what you know, but I know they're pretty good
https://youtu.be/Y09iNxx5nFE

and they'll keep getting better

[night girder]

Fair point, but one person can have multiple accounts 🙂

Depending on what the datasource is that got hacked/breached.

[cyan crescent]

Looks like AMDs new CPUs are kinda underwhelming

[gilded helm]

Zen 5 is probably going to be really good in laptops?

[willow pike]

strix point uses it and does seem to do real good

more graphs

[wanton orchid]

it do be do be do

[languid gulch]

in about 500 years the original music video is gonna be gone, but that meme will be found like a lost hieroglyphic

[sharp oasis]

Anyone know how to disable amd fan control

[twin dew]

Unfortunately that doesn't hold true for at least the early Intel AVX512 implementations (and not supported at all on desktop anymore).
Where first 50k CPU cycles ran AVX512 at very slow speed, and after that the max clock frequency took a very big hit...
So mixed use of AVX512 and other stuff caused performance penalty, not uplift.

In turn AMD AVX512 doesn't have such limits.

So opportunistically using AVX512 instructions on AMD Zen 4 and Zen 5 doesn't have any real downsides.

Also Intel first "implemented" (in microcode) AVX512-VP2INTERSECT command, that took 25 cycles.
Then someone found a way to do the same thing in software faster.
Intel later removed that command from supported ones.
Zen 5 implements it in HW with 1 cycle...

But the AVX512 stuff also is much better than GPU if your code has branches or interdependencies.
Or just isn't wide enough for GPU.

[twin dew]

"AMD fan control"?
Or MB fan control?

[sharp oasis]

Amd GPU fan control

[twin dew]

Why?

It should be running by the card firmware, unless you specifically enable overrides in Radeon Software, Performance, Tuning.

[sharp oasis]

I'd like to see if I can go close to passive cooling

Or more indirect cooling were I'll use case fans to cool it

And the fan software I use gets overridden by amd driver can control

[twin dew]

Not possible unless the specific card has native Zero RPM mode.
Or you can enable that in the Tunings.

[sharp oasis]

It does have zero rpm

[twin dew]

Not by AMD, by the card firmware not allowing "unsafe" stuff.

As in not by AMD software.
And the firmware cannot be overridden by design.

[sharp oasis]

Damn

So firmware is getting angry and saying nuh uh

[twin dew]

Not wanting the user to break the card by doing stupid shit.

[edgy hazel]

Unplug fan

[sharp oasis]

Sadly then the firmware gets mad

And puts the card into limited mode

Where it will run no faster then 500mhz on core and mem

[edgy hazel]

That would make water cooling the card impossible

I have no fan plugged into my card

[sharp oasis]

This card is very unique

In that it has too many quirks

It's been able to self reset SSPTs

Uhh self undo firmware changes that I applied.

Don't ask me how that's even possible I have no clue

And it says. I do what I want when I want it

[languid gulch]

lol

This is my GPU. There are many like it, but this one is mine. This GPU is my best friend. It is my life.

[dire igloo]

Not with CCX latency like this

Sauce: https://youtu.be/Th0LONlA9a8

[gilded helm]

Mobile uses mostly monolithic versions though.

I haven't looked at Strix point reviews, but I would hope they're good based on this.

[twin dew]

That was specifically from those, those have one CCX with 4 Zen5 cores, and one CCX with 8 Zen5c cores.

In monolithic die.

[wanton orchid]

cycles or ns ?

because if it's cycles it's misleading

[twin dew]

Usually ns for that type of charts IIRC.
The cross-CCX latency is huge.

This is explicitly in ns. for comparison

For Zen4 desktop.

[dire igloo]

6:35 in the video, nanoseconds

[twin dew]

Same test from Anandtech

www.anandtech.com/show/21485/the-amd-ryzen-ai-hx-370-review/3

But the point of that split is to have the actual stuff on the Zen5 cores with large L3 cache, and background stuff on the Zen5c cores with not a lot of L3 per core.

With hopefully no cross CCX stuff happening.

The Zen5c CCX has very little L3 per core, IIRC 4 times less per core than on the Zen5 CCX.

[wanton orchid]

as long as it stays under 100ns
it stays transparent for remotely correctly designed software
as for ccx under 500 is completely manageable
what you want to check is cache access latencies
no good software is significantly amount of time continuously tied to other thread directly in processing
also same ccx latency are good
and you don't want bigger main than 8 core on mobile most of the time
and again, ccx to ccx latency you should pay cost only some of the time, not every core to every core
the graph is as it is for showing all the information, but keep in mind this is not an area amount quality
I mean more red area does not mean more bad cpu as long as the green cluster are significantly big enough and at least mostly centered on the diagonal line

[twin dew]

If the scheduler works properly, there is very little need for the cross-CCX accesses.

And yes, 16MB L3 for the 4 Zen5 cores (4MB per core), 8MB L3 for the 8 Zen5C cores (1MB per core).

[wanton orchid]

and even then
you mostly pay for cache flush/recover more than for cross ccx latencies
only few process are tie to core to core bus latency without also relying on L3 parallel access/transfer

and most of these process should mostly be put on same core for L1 cache level syncing

[wanton orchid]

that is the second tier diagonal line latencies shown there

the '5-6.5'

[twin dew]

That is just inside same core between the core and HT-"core".

[wanton orchid]

that's what I said

[wanton orchid]

^

[twin dew]

Ah, sorry, misunderstood

Didn't remember the previous message anymore.

Doing too many things at same time...

[twin dew]

Anandtech isn't that good anymore.
All the real people who used to be behind it have left during the years after it was sold to large corp.

[wanton orchid]

tbh, these charts should come with reading guidelines
with at least the 3 main point I explained
I would want a visually amphasis on the diagonal line
like saturation levels could be good visually hinting
the 3 main point is :

• under 100ns is transparent, 500ns for ccx is still completely fine
• this is not cache latency, what people usually think of when imagining actual software execution properties is in fact shared cache access latency, not core to core cache coherency latency (which these graph show)
• everything in real life will focus on where the latency is low, as long as it is easy to do enough and large enough in area : the performance important part is directly centered on the diagonal line, everything far from the diagonal line (either vertically or horizontally but not both) will only be experienced few in real life software runtime

[twin dew]

Just chart to show how important NUMA awareness is.
Not constantly moving stuff between the separate cache domains because of bad scheduler moving threads around or by having interdependant threads running on multiple domains.

Neither which should happen anymore in most cases.

[wanton orchid]

yep
with mm and mmio performance counters
it should be fairly easy to spot these misscheduling

(and correct them in less than a ms)

I don't think cpu interface is able to hint the area of misses
it would be very nice information if it could (and is realistically possible to do)

but it's still only a matter of time for the scheduler to spot what thread is misbehaving with what other thread

tbh I feel good being able to talk with such clarity, confidence and detail
it really shows this is still my main skillset

(i.e not like when talking about ram timing configuration, cpu curve things, pcb design and 3d printing or other stuff like that)

[edgy hazel]

Ouch

ryzen 9000 delid gone wrong

[pure karma]

yea they are

there power efficient but at what cost

[twin dew]

Just enable PBO?

[pure karma]

auto overclocking isent a excuse for underwelming defaults tho....

[twin dew]

That isn't auto overclocking, that just changes the max power and current limits...

Same as if the CPU had higher stock limits.

Same as the ECO modes just lower those, as if the CPU had lower stock limits.

So before the "OC" mode was the default, now the "ECO" mode is default.

OC as in as high power limits as your cooler can support.

[dire igloo]

I doubt they are that power efficient compared to 65W Zen 4

[twin dew]

Ah, the GN efficiency chart didn't include the Zen4 ECO mode results for the 7950X...

[dire igloo]

It also excluded the entirety of Zen 4 non-X

So it compared what is essentially 65W TDP Zen 5 to 105W TDP Zen 4

And concluded that 65W TDP Zen 5 was more power efficient

[pure karma]

honestly the only reasson i ever remember pbo's existance is from it making the original threadripper's draw like 1000W without a care in the world

[dire igloo]

If only Zen 4 had very power efficient CPUs too

[willow pike]

eco mode my beloved

[rustic panther]

honestly it's nice to see a cpu manufacturer manage to get a generational performance improvement without throwing 400w at the silicon to manage it

[dire igloo]

Are we talking about the same CPUs?

[wanton orchid]

if it has same performance at lower power envelope it is strictly more efficient

[dire igloo]

Point is: GN didn't make it a fair competition by excluding those CPUs that'd be the strongest competition.

There was no apples to apples, 65W to 65W comparison.
It was 65W Zen 5 vs 105W Zen 4. Two variables changed, not one.
Their verdict is meaningless as long as 65W Zen 4 isn't being retested in the same environment

[wanton orchid]

your argument is pointless against mine
I repeat
same performance at lower power envelope is strictly more efficient.
you can't go around it, lower the zen4 power and you'll get less performance for same wattage
increase the zen5 wattage and you get even more performance

the 2 points are in opposite direction
so either way it's better efficency wise

your argument would hold if either :
zen5 was lower performance (even slightly)
zen5 was higher power consumption (even slightly)

higher performance and lower power consumption
IS
better efficency

now you could arg maybe zen4 could catchup at even lower power consumption
but it would be odd the tendency reverses

and then it's not 65 to 65 anymore
but 48 vs 48
or 30 vs 30

[willow pike]

fundamentally, amd should have made 65W(88W) the default for 7000 in the first place, and allowed the CPU to change power target based on user preference within Windows; say, using the balanced / best performance slider. then the user could get efficiency or maxed out performance as they like and we wouldn't be having this discussion

putting the switch for power targets deep in the BIOS, or even within ryzen master, isn't really good enough

anyone reading this won't have any problem fucking with such settings but most users would have a heart attack looking at the bios

like, I can change the power target of my steam deck to whatever wattage I want by pressing one button, and I can do it in game!

there's no reason gaming PCs don't deserve such an easy to access option

[wanton orchid]

they took the only occasion they could have to go back to eco default : having at least same performance doing so
and imo that's good

[dire igloo]

Unless we exclude margin of error

[wanton orchid]

the margin of error remains whatever the settings, and in either direction

[dire igloo]

Thing is: we don't know the margin by how much Zen 5 is more efficient because we lack an apples to apples comparison

You have to hold power efficiency against increase in cost to know when the better power efficiency pays for itself

[wanton orchid]

this is more of a safeguard imo
bios can still not follow guidelines
and OS can still require less performance to the cpu at higher targets
iirc amd cpu power targets does not change powers draw behavior at low requested operating states

[dire igloo]

Completely detached from the power efficiency argument:
How does Zen 4 versus Zen 5 perform with PBO enabled?

[wanton orchid]

People experienced huge margin with pbo on zen5 thanks to the lower default
zen4 was going from spec limit to ~actual limits
so expected bigger jump in zen5 (and so more performance too)
people said here earlier that it could get +25% from pbo
while afaik zen4 was significantly lower jump than that

more like +15% max

(which is still good)

and I've been running pbo enabled 5900X since 3 years now 24/7
so pbo is looking mature and safe to me

not like Intel already falling at stock settings

[dire igloo]

I gotta check if I can run PBO on my 3800xt

[wanton orchid]

didn't hear lot of people complaining about cpu failing under pbo in long run and it been at least 5 years now people have been using it

pbo is like : going from cpu model default limits
to ram/motherboard set + specific cpu unit set limits

which is why people have been doing oc from the start

So pbo is nice

it's the cheap safe enough one button "oc"

[languid gulch]

my 1700X refused to boost past its advertised 3.8GHz, but you know what? at least it would hit it and not melt

[dire igloo]

My 1700 was the first time I ever encountered an IMC causing instability

[visual tree]

Can't believe I had respect for this guy in the past. He has clearly lost his mind in recent years https://www.reddit.com/r/RealTesla/comments/1eln3pr/elon_musks_x_sues_advertisers_over_alleged/

Good luck with the lawsuit though. There is no way he is going to win this lol

[dire igloo]

You got any close up pics?

[edgy hazel]

Nah

[stray badger]

Marketing is important tho, if amd is going to say that the 9700x is an upgrade over 7700x, and it does not offer any stock performance improvements, that is not a great thing for people who don't know how to use OC and raise power limits

[twin dew]

After long and hard fight, got the DDNS updating to work...
Needed to update to newer version of the client than what was in Debian Stable, as the specific version was broken for the specific protocol...

[night girder]

YT testing out fact check function.

https://support.google.com/youtube/answer/14925346

[edgy hazel]

who wants to write shit on mobile?ß

[night girder]

To help others stay informed, you can sign up to submit notes on videos you find inaccurate or unclear. After submission, your note is reviewed and rated by others. If your note is rated as helpful by many people with a broad set of perspectives, it may be published and appear below the video.

[night girder]

It's a test, not released feature. Still I think it can have value. More than other features they added.

[edgy hazel]

regex should just be illegal by now

The selection of data in the
channel file was done manually and included a regex wildcard matching criterion in the 21st
field for all Template Instances, meaning that execution of these tests during development
and release builds did not expose the latent out-of-bounds read in the Content Interpreter
when provided with 20 rather than 21 inputs.

[languid gulch]

you're insulting the competence of all 3

[languid gulch]

stupid question time

do spinning drives always have "active" power going to them?

like, do they always have the amount of power going to them that they use when they're spinning up and actively being used for file transfers

[mental oriole]

Mobile only, and US only... hah...

[soft bloom]

they can be shut down entirely

or do you mean "spinning up for active use"?

[night girder]

Ah, I missed that part.

[languid gulch]

i'm wondering if keeping them powered on "fully" makes them more responsive

[soft bloom]

"responsive" is more about good cache

[gilded helm]

I believe HDDs do exist which power off the spindle fully. Can't comment beyond that/

[modest isle]

i thought he was supposed to be electrifying America, but the past few years he has gotten himself into a seemingly endless string of battles over social issues. why couldn't he just keep his head down and build things like he was before???

[languid gulch]

i think he looked at how much all of that would actually cost and realized he was too greedy to actually do it

[stray badger]

I think he figures that fewer regulations will help him in the long run

And that by supporting the traditionally anti-ev people, they will support him and tesla more, and since the left likes evs, he is ensuring that he will not be stopped from making EVs by either side i guess

More likely, he is just mean

[wanton orchid]

yes
very low power state does not even run the platers

So it have to spin up and ready heads all the way on the first little request

[twin dew]

You can even query drives for their different power states and how long it should take for them to return from them to full operation.

And should also show how long after previous access from OS they will go to each of them.

But I might be mixing NVMe and SATA with that.

[soft bloom]

they all can. it's just a setting

[twin dew]

Some did that so excessively they broke down in short order (Some versions of WD Green series etc.)

[soft bloom]

i think cache can give response without actual physical action

[soft bloom]

why do you say about it like it's something dictated by drive itself? i was configuring this setting on both windows and ubuntu

[twin dew]

Not for any real data.
And often even SMART query spins up the drive, even when no actual data is asked, just firmware SMART status

[soft bloom]

i mean, i assume there's the default set by manufacturer...

[twin dew]

That is when the OS asks the drive to do thing.
But they also have internal timers.

Those WD Greens were parking their heads in less than a minute.

And there is limited number of life in that mechanism in parking and unparking cycles.

In several hundred thousand cycles range, IIRC 600k for the Seagate Ironwolf Pro NT:s.

And the platform firmware can also do PM stuff, bypassing the OS.

[soft bloom]

oh, it's a bit more complicated than i remembered:
standby - on/off never..3hours
APM level - on/off savepower..performbetter
write cache (unrelated to this topic, but also configurable) - on/off

[twin dew]

And seems that that table of the latencies etc. is just NVMe thing, not in SATA yet.

Note that there is no get option because ATA standards do not specify a method to read the standby timer.
🥲

[soft bloom]

dam, i can't update my ubuntu laptop because /boot volume of 512MB is too small 😦

it wants at least 140MB more

when did kernels get so large?

[twin dew]

Hmm?
Something weird is going on

Kernel is 7,8MB and initrd is 8,4MB for me.

Then other stuff including the contents of two EFI partitions and the GRUB for total of 53MB with just one kernel.

51M total on another machine with two kernels and one EFI.

35M after I removed that older kernel as unneeded.

I would expect you don't have any compression for the initrd going on and you include all modules, not just the needed ones?

Because the whole raw module directory for one kernel is almost 400MB for me.

/etc/initramfs-tools/initramfs.conf or new file in /etc/initramfs-tools/conf.d/ with "MODULES=dep" instead of "MODULED=most"

Hmm.. message got content blocked...

And COMPRESS=<some supported compression>

# COMPRESS: [ gzip | bzip2 | lz4 | lzma | lzop | xz | zstd ]
COMPRESS=zstd

Need the packages for the specific compression installed.

zstd for Debian for that zstd, which is the Debian default.

Again, the directories were Debian specific, but Ubuntu should use the same.
Non-Debian based distros have the initrd generation configs in diffent places.

After editing, run "update-initramfs -u" to update the current initrd to the new settings.

Just setting the compression first, and only setting the smaller module selection after the kernel update would give way to run the older kernel with larger initrd if all the necessary modules aren't included automatically.

The probe is mostly complete, but for me it doesn't load xxhash_generic for BTRFS non-default hash by default as example.

[wanton orchid]

entreprise grade 20TB disk have 512mo of cache, and it is mainly used for write not for read, now do your statistics with that
also, if you query some data, it's activity, you will likely query some other data too, so you HAVE to spin up the drive all the way again

other wise you would need a special interface, for OS to hint, but again that would be only opportunistic for rare senarios

i.e absurd

a disk is a disk, you WILL go to the disk

[twin dew]

And took second look into XXHash64 vs. HW-accelerated SHA256.
XXHASH is at 870 cycles or so per 4KB, SHA-NI SHA256 on Zen3 is at around 7000 cycles for same.

And because I'm not doing deduping, the 64 fast non-crypto-rated algo is perfectly aquate and 256bit crypto algo isn't really needed.

CRC32C is the default, 32bit hash with 470 cycles per 4KB.

[twin dew]

For the BTRFS checksum hash function

[wanton orchid]

nvme have this data directly published to OS for sure (in spec)
but I dont know for sata, some information may be available too
but iirc platter spinup time vs performance target value is often missing
but very easy for OS to pinpoint by simply testing

[twin dew]

That was the thing, I checked after that line and smartctl or hdparm neither allowed viewing something like that for SATA.
So it was just NVMe thing.

To have such queriable table about the different sleep states.

[twin dew]

And SATA doesn't even contain way to view the current spindown target value, which this was referencing, the quote is from either smartctl or hdparm MAN page.

So it allowed to set the standby timer to off (if supported), and between seconds and 5.5h (again depending on actual drive implementation).
But not getting the current value.

[wanton orchid]

but hdd usually have very predictable behavior, so it's not really required in practice, that's probably why it never was added to the interface

[twin dew]

Would have been nice to know if setting it is required to get sane behavior, or if the default works.

As the 3TB Reds are at 100k and 110k head load cycles at this point.

Need to check the values again in few days.

[wanton orchid]

I mean the OS can practically be aware of that

each every of these 100k, OS experienced the big stall

[twin dew]

The 100k includes shutdowns etc.
And as those were never used as the OS drive, but as storage even when new.

Linux doesn't touch that stuff at all if you don't personally configure it to on most distros.

[wanton orchid]

but then that's linux choice, and likewise linux could report the stall duration it experienced

or you could even test it directly through kernel with performant enough script

like :
wait_for_spindown()
timer = time.current()
open(hdd_file).read()
spinup_time = time.current()-timer

[twin dew]

Don't think even that is needed, as there seems to be some way to view the power state of the SATA stuff already, as you can limit Smartd to not query drives that are sleeping, to not cause spinup.

[wanton orchid]

I still was talking about the original topic of "time to power state table"
for having the time to spin, you have to test spin it

but then it will be very reliable

so it is easy to query

you could create a powerstate table tracing script

hdd are mostly either up or down

so it is easy to test

[twin dew]

Fully operational.
Head parked but spindle still spinning. (on some drives, switches between this and full was what broke those overaggressive greens)
Spindle in slower speed. (on some drives)
Spindle stopped.
SATA link in power managed mode.

The specific greens were doing it with less than 2 minute timer, don't remember the exact seconds, without care of that main standby timer.

Needed wdidle3 (specific utility), or now included in hdparm, to tune that.

[soft bloom]

Ubuntu told me there's like 3 kernel versions there, and each is around 100mb but new update needs 140mb more free.
This is second time this summer

[twin dew]

Ah, 8 seconds as default.

[twin dew]

The kernel, or the initrd?

ls -lh /boot

[soft bloom]

I definitely remember that windows can display first level of folder without actual spinup, and when i go further it starts spinning. Don't have such memory for Ubuntu, maybe because i told it to not spindown cayse i use it differently from there

[wanton orchid]

kernel should be 15MB~ at most

[twin dew]

Expectation is that the initrd:s aren't compressed.

[soft bloom]

Why i think it had kernel version?...

[wanton orchid]

that simply does not go to hdd at all, windows just cache directory structure

[soft bloom]

Oh btw yes, correct - it suggested compressing them

[twin dew]

As the full module directory for my Debian kernel is almost 400MB.

Not using compressed initrams is just stupid.

In most cases.

I put lot of information earlier.

so uncompressed initrd with "most" modules is gigantic.

[soft bloom]

I expected cache to be used for read access as well. It's not like anything can be done with data without cache being able to notice that. I don't want data loss so i disabled write cache

[soft bloom]

Why it's not default then?

[twin dew]

It is in most cases?
No idea why your system doesn't have it on.

[wanton orchid]

write cache does not do any dataloss if the OS is not dumb

[twin dew]

Debian doesn't even ask, it just does it.

[soft bloom]

Why not just keep dir structure in the cache? At least realisitic amount.

[wanton orchid]

also you are mixing various cache here, I said windows is caching directory structure
this have NOTHING to do with the hdd, it will not even send notfication on the sata line at all

[wanton orchid]

it's not hdd job

[twin dew]

HDD has no idea about file systems, it just stores blocks of data.

[soft bloom]

Ubuntu warns about it. I think its straight up about potential of loosing data is power is lost

[wanton orchid]

it does not

[soft bloom]

Ok fair

[wanton orchid]

let say you write some text file, in text editor

your pc then crash

would you rather have the data tried to write, or not ?
in both cases it is wiped

[soft bloom]

I have to say that windows / explorer is dumb

[twin dew]

There is very small chance that write that has been sent, but not actually written, can be lost on power loss.
And some drives to lie about having actually flushed the data to the drive from the drive cache.
Sending the "Ok, written" as soon as it is in the cache.

[soft bloom]

I remember something like that...

[twin dew]

And then not being able to actually write it if the power to the drive is lost in just the right moment, allowing even synced writes to be lost.

Linux just warns to the extreme about the issue.

[wanton orchid]

lier should be detected, it is a known issue, linux kernel have long safeguard list of things against that, as well as blacklist

[twin dew]

It does exist, but isn't that realistic issue for most systems.

But the warnings remain.

[wanton orchid]

again ^

[twin dew]

In most cases it doesn't even matter, like that shows.
Just that in some cases journaling file systems lose their "protections" if the stuff they do happens in wrong order and the syncs don't happen right.

[wanton orchid]

the write ok feedback, is not a data-on-the-fly problem, but an hdd acknowledgement / consistency problem (which should be addressed in not dumb OS)

[twin dew]

Disabling the write cache just lowers the window of opportunity for non-synced writes, or writes where the drive lied.
But also trashes the write speeds.

[wanton orchid]

anyway it should also use the OS backed write cache

it is a pipeline

[twin dew]

Yup.
Normal file systems can get "trashed" if suitable write is missed.
Metadata journaling file systems (NTFS, EXT3, EXT4 etc.) can lose data, but the file system itself will survive unless those syncs are lied about.
Full journaling/Copy-on-Write file systems (ZFS, BTRFS etc.) should have one or the other data still in even after such failure, again if the syncs aren't lied about.

[soft bloom]

This conversation makes me think that i was write about switching place to store enormous http cache from hdd to ssd. But also i could use LVM caching. But maybe even better - use (extremely large) sqlite? Nah i don't believe it can handle 100GB

[twin dew]

So even with ZFS, you can lose the transactions ongoing when the power loss happens, but you will still have consistent file system from just before that transaction.

[soft bloom]

I think my next ubuntu installation will be on zfs

[twin dew]

If the drive didn't lie about synchronous writes.

[soft bloom]

At least if i don't lose anything by that

[wanton orchid]

tbh lvm caching does not look that good at all

[soft bloom]

My long history of OS's with backups has collected some stupid setting over the years

[soft bloom]

That's better than nothing. Especially when i just reuse decade old ssd

[wanton orchid]

depends

you add a failing thing on the chain

read cache trashing a something to be afraid of

always do backups of the most important data BEFORE unmounting the cache
because if the cache is not very consistent, it can hide filesystem issues

[twin dew]

But there are reasons why I used BTRFS in DUB-mode for both metadata and data with xxhash64 for the Pi SD-cards.
DUB: every block is written in two places, kind of RAID1 on one device, doesn't protect against full device failure, but allows for cross-checking and only one of the blocks needs to match the checksum for recovery.

BTRFS uses DUB for metadata by default on single partition/device filesystems, but data is just single by default in that case.

Mainly in my case for bit rot detection and correction for long time filesystems.

https://www.youtube.com/watch?v=YKbDApzT1iw

[languid gulch]

i've been hearing mixed things as to whether or not starliner can be brought home autonomously

which to me should instantly disqualify it from future flights & contracts

[twin dew]

Boeing didn't fully test that code and it might even have been stripped out of the build for the capsule on the station with expectation it would never be needed with astronauts on board to take over.
To save costs.

"This was just needed for the first test flight"

[languid gulch]

i feel like boeing is trying to stretch it out as long as possible to avoid consequences for the company

[twin dew]

So the interactions for that part of the codebase to any changes since the second unmanned flight have not been tested.

[languid gulch]

the longer it's up there, the further out investigations & punishments will be

if i was the US govt i'd be suing them for fraud

and endangerment

i think youtube is starting to crap itself. when i click on a video, it'll go to some random timestamp

[night girder]

YT got blocked today in 🇷🇺

[languid gulch]

i think you mean russian occupied ukraine

[night girder]

No, Russia 😄

www.msn.com/en-us/news/technology/russia-is-blocking-youtube-by-throttling-internet-speed/ar-AA1ou0Lu

[twin dew]

Russian occupied trans-russian-states.

[night girder]

but YT working perfectly here.

https://tenor.com/view/it-turning-off-gif-19018125

[jovial dirge]

https://tenor.com/view/hello-old-people-computer-technology-biff-gif-12085559

[languid gulch]

or as i saw it yesterday, the Kursk People's Republic 🤣

[night girder]

#copium

[twin dew]

*Whee*
https://www.amd.com/en/resources/product-security/bulletin/amd-sb-7014.html
Sinkclose named long time AMD CPU bug found, more details tomorrow?

[languid gulch]

lol

yea but it's yet another one of those "if they can get to it, they already have access to everything else"

[night girder]

You read about slubstick baldur? I posted about it. Forgot if you responded to it or not.

www.stefangast.eu/papers/slubstick.pdf

[twin dew]

Just Wired article out at this time and then copies of that, the actual people should be doing presentation tomorrow at Defcon

[night girder]

About slubstick? I am confused. I just posted the thesis.

[twin dew]

About that Sinkclose

That the only real information is that AMD bulletin, and crap Wired article, and then rehashes of that Wired article.
With more details tomorrow.

[night girder]

/linuxsecurity.com/news/security-vulnerabilities/anatomy-of-slubstick-linux-vulnerability

Security Awareness: Businesses should ensure staff members know potential security threats and the significance of maintaining robust security hygiene.

How are staff to know about security threats, when even scientists don't know about them and discover them 😄

But yeah, not sure how a user can protect memory.

This is pretty low level stuff.

[twin dew]

Doesn't seem to have CVE assigned.
As that one is much more theoretical and needs other stuff to be usable.

Like the already fixed other security faults they reference.

[night girder]

It's an enabler.

[twin dew]

Yup.

And back to that AMD thing...
Again shows their priorities.
All the Epyc CPUs have firmwares available that fix the issue, since 2024-04-04 for earliest and 2024-07-11 for latest.
All the embedded CPUs have target of October 2024.
Ryzen 3000 will not even get a patch.
For the 4000 and up, the AGESA versions with fixes are in theory out in just last few days, but it will take ages for them to go through the MB makers.

[night girder]

Sinkclose is already fixed before the report is out?

[twin dew]

That is the norm.
That the researchers who find issue contact the manufacturer, and give time to develop a fix and get ready to deploy it (or even deploy with closed source stuff) before the fault is publishes.
If the manufacturer plays ball and doesn't drag it out.

[night girder]

Not always it seems 😉

Some researchers make it public to add extra pressure for fix. (a.k.a. slubstick).

[twin dew]

But yes, for server platforms, the fixes should be available.
For embedded, not available for few months.
For consumer platforms, when MB makers get aroudn to it.

Norm, not completely universal.

[night girder]

Maybe they get a week or two heads-up. But not all researchers are that polite. It's not always done.

[twin dew]

Even Google is asshole and has specific pretty short window for all cases before disclosure.

[night girder]

I remember a few stories of researchers communicating with company. Company doing nothing about it. Researchers publish exploit public. Company "suddenly" has a patch.

[twin dew]

Yes, that was the comment about "if the company plays ball and doesn't drag it out"

[night girder]

I can understand that maybe Linux is more "public" property.

Or community driven, so more communication with community too.

[twin dew]

With open source, the fix is usually developed behind closed doors, and all the major distros security people get heads up, and everyone pushes the change to production at same moment the disclosure is made.

[night girder]

Yeah, something like that.

[twin dew]

most major parts of open source community have specific security issue reporting ways that don't push the info to public view.

[night girder]

I know a few people who did bug bounties. It can be very frustrating.

If you spend a lot of time in finding and documenting bugs/exploits, and you believe it's critical and then they don't accept your report so they don't have to pay you. It sucks.

Still to this day, I believe major IT companies are both quick and slow at the same time.

And with that I mean, some companies are so sluggish/slow to respond to security risks. But if they want, it can be patched in a day. (slow/quick).

techcrunch.com/2024/08/08/security-bugs-in-ransomware-leak-sites-helped-save-six-companies-from-paying-hefty-ransoms/ 🤣

Some of the bugs include the Everest ransomware gang using a default password for accessing its back-end SQL databases, and exposing its file directories, and exposed API endpoints that revealed the targets of the BlackCat ransomware gang’s attacks while in progress.

Stykas said he also used one bug, known as an insecure direct object reference, or IDOR, to cycle through all of the chat messages of a Mallox ransomware administrator, which contained two decryption keys that Stykas then shared with the affected companies.

[gilded helm]

What's with the trend of banking websites disabling autofill functions and breaking password managers? Is there an actual security justification for this?

[gilded helm]

Surely forcing people to copy/paste just switches the vulnerability to clipboard sniffers? Or worse, password reuse? But I guess that just let's them shift blame? 🤷

[night girder]

you can still copy/paste?

[gilded helm]

Yep

[night girder]

Because that's something some websites disable out of security.

[gilded helm]

Autofill functions are disabled, but not c/p

[night girder]

So, you can copy/paste. But passwords managers can't interact with the input fields?

[gilded helm]

Right

[night girder]

Let's hope that's a bug?

[gilded helm]

Yeah, doesn't seem sensible.

[night girder]

Well, either block both. Or allow both imo.

Because if it's for security, better do it properly.

[gilded helm]

And I have my doubts that disabling them encourages a good security environment, that'd definitely lead to bad passwords and/or reuse.

[night girder]

I trust the auto-fill feature more than copy/paste tbh.

[gilded helm]

Or worse, people installing third party extensions to "fix" banking websites. Definitely no security concerns there.

[night girder]

Have you tried different browsers?

[gilded helm]

Yep

I'm on IronVest, their support confirmed it's a Chase issue, and Reddit has complaints about it with other financial institutions and other password managers spanning years.

[soft bloom]

How many of you watched Bugs Life and copied Flik to be IT guy who contributes by doing some automation/invention etc?

[night girder]

niice

[edgy hazel]

holy shit this is def still alpha software

lmao

[night girder]

software?

[edgy hazel]

isn't an OS software?

[night girder]

Yeah, but we call it OS

[edgy hazel]

OKAY

[night girder]

🤣

[edgy hazel]

holy shit this is def still alpha OS

[night girder]

Buggy? Or not many features?

[edgy hazel]

it's quick but there are still some bugs

[night girder]

Is this information true?

Initial release October 27, 2017; 6 years ago[1]
Latest release 22.04 LTS[2] Edit this on Wikidata[3] / 25 April 2022; 2 years ago

[edgy hazel]

popos 24.04 got released into alpha yesterday

[night girder]

2 years without updates...

[edgy hazel]

it's LTS...

[night girder]

What about security updates?

[edgy hazel]

ofc it still has security updates

LTS is "Long Term Support"

[night girder]

Was just wondering how they handled it.

[edgy hazel]

like ubuntu

the "2 years" just means they didn't release a new version of it

[night girder]

Why did you pick it?

[edgy hazel]

over something like ubuntu? It has more support for gaming stuff out of the box

[night girder]

and does it game well?

[edgy hazel]

basically everything except riot games games

because their anticheat forces you to use windows

which tbh is pretty good for my mental health

[night girder]

depends on what other games you play 😛

but yeah, if you can live with that. Why not.

What do you miss out on? League of Legends?

[edgy hazel]

yup and tbh I didn't play it much anyways

[night girder]

You can always try Dota 2 😉

[edgy hazel]

pfft

[edgy hazel]

"Congratulations on getting rid of the flu, would you like to try a tumor next?"

[languid gulch]

wtf is Yarn, and why has it taken over my gif search results, and how can i stop that

[edgy hazel]

https://tenor.com/view/little-big-planet-angry-sackboy-gif-120402395630872294

[languid gulch]

why does tenor for gifs not show up anymore for google

i don't get it

the yarn links are always broken

[pure karma]

yarn?

for like knitting?

[languid gulch]

gif site yarn

i'm not getting any tenor or giphy links when i'm looking up gifs for memes

just this yarn shit

[pure karma]

starting to get real tired of either my headphones or headphone wire but holy hell hearing constant static and poping is the peak of anoying

its probably the wire because anytime i move the wire it stops

[languid gulch]

relax, it's just the brain worm RF interference

[sharp oasis]

Gotta give it to my ranger. Saved my life. (I hit a moose)

[visual tree]

Ouch, my mother hit a deer while leaving a tunnel. People told her to leave a dead deer on the side of the road and don't report the incident to the police.

We have really dumb laws here and you shouldn't be fined if you hit a wild animal if you drive under the speed limit

But people still get fined if you drive under a speed limit and hit a wild animal because there was a wild animal sign and you "should have been more careful"

Like, do people who make dumb laws realize even if you drive very slowly, a deer hiding in the bush could jump out and you don't have enough time to react....

[languid gulch]

glad you're ok, truck did its job well

[dire igloo]

Old Canadian wisdom:
You see a deer, you keep driving. Deer is soft, deer gives in.
You see a moose, you stop or swerve. Moose is solid, moose doesn't budge.

[tribal kraken]

New beta BIOS out for my MB

[edgy hazel]

About to go back to 22.04

Alpha release being alpha

[tribal kraken]

Yeah, not bothering toi update now. If the CPU dies now the beta BIOS most likely was not in time to save it

And I think there was rumors for new gen of CPU to go on LGA1700 anyway

[twin dew]

Fuck the SystemD NetworkD DHCP-server SendOption parameter...
To send more complex options, they need to be set to be to type of string, and then use per-byte hex encoding with \x##...

The type takes one of "uint8", "uint16", "uint32", "ipv4address", "ipv6address", or "string".
And when you need to 5-9 bytes per entry, and I needed 6 entries...
For Classless Static Routes to be sent out by the DHCP-server

So take IPv4 subcomponents, convert each to two hex digits, add, redo for next.

[edgy hazel]

Aaaand we're back to 22.04

Hello stable OS

[deft sinew]

Anyone here using an rx 7700 for satisfactory? Looking into buying one and wondering how it performs

Probably in 1440p

[twin dew]

Radeon 6600 at full-HD gives 75-100 FPS on new save with maxed settings without lumen/global illumination, without upsampling.
And still 60+ FPS with over the normal limits save if your CPU is good enough, which needs AMD X3D or very highest end Intels, otherwise you have been CPU limited for a long time already.
So yes, good enough if you don't have very high end CPU.

More when looking at sky or ground, but as looking around in the worst performing biomes.

[twin dew]

Anyways, might have hopefully found the reason for my BSODs after the reinstall, antivirus fucking with install of AMD GPU kernel driver because older version had vulnerability, and not differentiating between vulnerable and fixed versions:
https://www.amd.com/en/resources/product-security/bulletin/amd-sb-6009.html
Real-time file system protection
F:\Baldur\AppData\Local\Temp\USBCPDFW\PDFWKRNL.sys
Win64/AMD.G potentially unsafe application
Event occurred on a new file created by the application: F:\AMD\AMD-Software-Installer\Bin64\7z.exe

Made exception to not use that detection and did DDU & Reinstall cycle, including setting DDU to remove that specific driver that isn't normally included in the cleanup.

Popped up right after OS reinstall, because I had installed ESET before installing AMD GPU drivers, but just allowed, but probably too late.
And hadn't popped up again on driver upgrades until now.

But we shall see...

But something running in kernel-mode legitly has been trashing kernel memory space to cause few specific variants of KERNEL_SECURITY_CHECK_FAILURE BSOD

[jagged snow]

https://arstechnica.com/security/2024/08/almost-unfixable-sinkclose-bug-affects-hundreds-of-millions-of-amd-cpus/

Wonderful

Not sure if it was already sent

[twin dew]

Talked about 24h ago.
But only unfixable if infected before installing new AGESA with fix.

As the active infection is unfixable, not the exploit to infect.

Ah, and that is the same clickbait Wired article from yesterday just reposted as is by Arstechnica.

[jagged snow]

Oh fun

Yeah, it doesn't seem quite as serious as they're making it out to be

Less of an issue than those insecure secure boot keys

[twin dew]

Basically Wired wrote lot of clickbait based on very partial release as the details are still under wraps until Defcon presentation by the finders.

[jagged snow]

Yeah...
Requires kernel level access to exploit anyways, so imo by the time this specific vulnerability is an issue you've got bigger fish to fry

[twin dew]

All the Epyc CPUs have firmwares available that fix the issue, since 2024-04-04 for earliest and 2024-07-11 for latest.
All the embedded CPUs have target of October 2024.
Ryzen 3000 will not even get a patch.
For the 4000 and up, the AGESA versions with fixes are in theory out in just last few days, but it will take ages for them to go through the MB makers.

1.2.0.cb for desktop AM4, 1.2.0.1 for desktop AM5.

And MS is almost certainly including the new firmware versions in Windows to be loaded into the CPU at boot time if the MB didn't do it already.

As soon as they finish testing for this specific release.
Older ones have been done for a long time already, so the basic method is solid and tested.

Wired and ArsTechnica are under same owner now, who likes to push crap Wired articled to ArsTechnica as is to get more clicks.

Also lot of other sites have just gone and regurgitated the Wired speculation since it was published yesterday, as only actual source at this time is that AMD security bulletin.

As repost: https://www.amd.com/en/resources/product-security/bulletin/amd-sb-7014.html

So the presentation will be today 19:30-20:15 UTC.

info.defcon.org/event/?id=54863

So that will give the actual details of which CPUs are known to be affected, what the finders think can be done with the exploit etc.

[night girder]

Doesn't sound that bad.

This silicon-level issue appears to have remained undetected for nearly two decades.

[twin dew]

Doesn't seem like that based on the AMD stuff, we shall see with the presentation.

How much of the Wired clickbait is just out of the writers head, how much is real.

[night girder]

It's from the presentation. The quote.

[twin dew]

When the presentation isn't out yet?

[night girder]

Read presentation info?

[twin dew]

With Wired?

[night girder]

And it's writen in "Our team noticed..." - form. So I assume it's the researchers writting that little intro.

[twin dew]

I will comment once the full thing is out, and hopefully with some AMD side comment too...

[night girder]

I mean, I doubt they will be lying about it.

Also they say: "appears to have remainded undected"

[twin dew]

it is just counter in seconds from specific moment.

[night girder]

Also seems the fixes are planned for october 😒 From AMD.

[twin dew]

Those are for the Embedded CPUs.

[night girder]

Oh right.

[twin dew]

Like I said, the Epycs have had the fixes out for long time, and the desktop platforms got them few days ago.

From 4000 up.

[night girder]

For a moment I was hoping Asrock being fast 🤣

[twin dew]

Also 3000 threadripper etc.
Different spread for laptop platforms.

But like I said, MS is going to make Windows install runtime ucode in early boot each boot.

[night girder]

source?

[twin dew]

That is common thing for long time already.

techcommunity.microsoft.com/t5/core-infrastructure-and-security/demystifying-microcode-updates-for-intel-and-amd-processors/ba-p/1000845

[night girder]

I have more questions now.

[twin dew]

Basically the CPUs can take signed microcode updates from OS.
Best done at very early boot.

Both Windows and Linux have the functionality to do that.

[night girder]

Modern processors do have the ability at initialization to apply volatile updates to move the processor to a newer microcode level. owever, as soon as the processor is rebooted, it reverts back to the microcode baked into their ROM.

Then what are BIOS updates? It updates the motherboard but also affects microcode from CPU's no?

[twin dew]

MB firmware contains the microcode at some level, deploys it to the CPU during POST.
Then OS can optionally change to newer one.

[night girder]

Does the motherboard also load the new microcode in everytime?

So, microcode in ROM = inmutable. When PC boots up, new microcode gets "injected" from Motherboard or OS to overrule that microcode?

[twin dew]

CPU doen't contain microcode at all AFAIK by default.
MB loads whatever microcode version is included in the current BIOS FLASH during CPU startup.

[night girder]

But if I put the CPU in a new motherboard, it's back to "basic" (the one that was baked into it?)

[twin dew]

As in doesn't contain any microcode anymore.

[night girder]

When processors are manufactured, they have a baseline microcode baked into their ROM. This microcode is immutable and cannot be changed after the processor is built.

[twin dew]

Ah, yeah, forgot, the ucode updates are just overlay patches.

Not full microcode.

If they were, having base ROM wouldn't be needed.

[night girder]

So, CPU has microcode baked into. Motherboard loads it in. If newer microcode is found (from BIOS updates or pushed from OS) then MB will use that microcode. Put CPU into new motherboard. Back to 1.0 microcode.

Sounds really inefficient.

[twin dew]

So CPU has that base ROM.
Then on boot MB loads the current ucode overlay from the BIOS to the CPU.
Then later the OS can optionally load newer ucode overlay to the CPU.
Both of the overlays get lost when CPU loses power.

And get reapplied on next boot again.

[night girder]

Yeah, ok.

[twin dew]

So you get at least the ucode version included in the MB BIOS you have.
As that gets applied by the MB during the POST.
And then can get higher level for security exploit reasons from OS made update during OS boot.

So you are protected in OS, once either MS deployes the ucode update for this issue via Windows Update.
Or once you get BIOS with the AMD listed AGESA version.

[night girder]

I hope they never find an exploit to block ucode deploys from MB to CPU.

Becasue then the whole secuirty would collapse, since that code contains the fixes.

[twin dew]

You can, but that needs physical access?

[night girder]

Yeah, but I mean, remotely.

[twin dew]

That is so scary about exploit like that, as they could in theory allow for such implant.

[night girder]

I never knew the BIOS updates were volatile.

[twin dew]

That is the point of the BIOS FLASH chip, which acts as the storage, from which the stuff is loaded each boot.

Or any other (EEP)ROM chip or like for the firmware.

[night girder]

I can see why it's also not feasible to have ~~RAM ~~ permanent storage instead of ROM on CPU.

[twin dew]

Would be way to kill the CPU.
As any of the current such techs have limited write endurange.

[night girder]

I know for a lot of people, low code stuff doesn't sound that scary. What can go wrong? And exploiters cannot simple do this and that. But to me, it sounds very scary. You want the foundation to be as solid as possible.

But, like you said. Let's wait for the presentation before panicking. Maybe you have to do some really stupid stuff to get the exploit to work.

[twin dew]

Or bad microcode update bricking the CPU.

[night girder]

There is a difference between theoretical and practical.

[night girder]

Yeah, and if you would put CPU into new motherboard, it would open some new issues. Since CPU still has code from previous motherboard that might not be compatible. So you have to reset that etc etc. Or find a work around.

I love Tweakers.net, translated to english for you Baldur:

AMD emphasizes that the vulnerability is very difficult to exploit, as hackers must already have access to the kernel of a system to do so. The manufacturer compares the bug to a method of opening the safes of a bank while the alarms, guards and safe door have already been bypassed.

You read that article I posted yesterday about researchers hacking the website of a hacking group and helping 6 companies decrypting their PCs with keys?

[twin dew]

Yes, but I really need to head to bed now.

[night girder]

Ok 👋

[twin dew]

Database of found ucode update files:
https://github.com/platomav/CPUMicrocodes

For the newest for each CPU.

As AMD doesn't distribute those themselves for linux use except for their Epyc line, and as ASRock Rack is way behind, I have had to use one from there for my server.

[night girder]

https://tenor.com/view/monsters-inc-sully-you-go-to-sleep-go-to-sleep-go-to-bed-gif-22295511

[twin dew]

So need to keep eye out for update with that new fix.

I was just browsing it for other reasons already, just realized it was relevant.

[night girder]

I was just joking mate 😄

[edgy hazel]

bro what are you working as? tmrw it's sunday and it's like 8pm rn

are you a baker?

[twin dew]

Just slept too little lately, it is not overly hot.

So trying to claw something back.

So aim is to get 12h of sleep.

[edgy hazel]

you do know sleep doesn't work like a mana bar right?

[twin dew]

Not for overcharge.
For being sleep deprived, somewhat.

[edgy hazel]

man this filament gotta be like 4 meters max

The Prusament PETG Tungsten 75% 100g sample is approximately 10 meters long.

so 1kg is 100m. normal PLA is like 350-400m

[dire igloo]

And it's tungsten infused or smth?

[edgy hazel]

Its 75% tungsten

[night girder]

You knew you have food safe 3D printeable materials? Because I didn't.

I was wondering, how safe would it be to print a 3D cup and drink from it? And if you use right printing material, it should be safe for your health.

[edgy hazel]

PETG is foodsafe and dishwasher safe. The issue is usually the microscopic gaps you get when 3D printing

Which is why you have to seal it with foodsafe epoxy if you ever want to use it for anything food related

[lethal pebble]

i love 165 fps cause its better

[safe trench]

not quite

for satisfactory lower fps is kinda better since hypertube cannon

[tough owl]

I got invited to play valves new game 👀

[safe trench]

what game?

[tough owl]

i got no idea if I can share anything

but just google valves new game

[sharp oasis]

Or hot the breaks as hard as you damn can

I was saved my losing half my speed before hitting the moose

[safe trench]

bro..

it was suposed to come out on 2025

not 2026

i cannot wait any longer

the sheer fact how paradox is using these tatics to stop people with gamepass to get the game for free is insane

and how both playstation and xbox are paying then tens of thosaunds if not hundreds of thousands

should things like this even be legal?

2025 is more than enough time to at least make a stripped down verison of the game

[night girder]

You can wait.

[safe trench]

i cannot

this is tourture

[night girder]

It's buggy and needs more work.

[safe trench]

why realse an unfinished game then?

[dire igloo]

https://tenor.com/view/money-mr-krabs-gif-18326632

[night girder]

💰

[lethal pebble]

wow that sucks, i played it so much its awsome idk why people say its glitchy

[languid gulch]

oh just watch the reviewer release vids, you'll see why 🤣

[willow pike]

it wasn't done, unity shafted them hard by not finishing half of the features CO were promised they'd get, forcing CO to implement a bunch of core features themselves. the only reason it got released on PC is that paradox demanded money for that quarter. it's still not finished, and performance is not good enough for consoles

[cosmic imp]

Trust me you can lol, the CS2 needs a LOT of work before it becomes fun just stick with CS for now

As far as I know for some reason they decided to go with the relaxed city builder where you don't really need much involvement

They're just recently addressing econ

[safe trench]

im gonna steal your pfp

[cosmic imp]

Wtf

[safe trench]

its in my folders now

its too late

[cosmic imp]

Someone stop him

MODS PLEASE

[safe trench]

hehawhaw

[cosmic imp]

He can't keep getting away with this...

[safe trench]

time to steal your user

[cosmic imp]

WHAT

[mental oriole]

[willow pike]

[visual tree]

https://youtube.com/shorts/mAvWBub8Ycc?feature=shared

I'm not a software engineer but I already have enough work experience so I don't have to deal with this bullshit anymore lol

[twin dew]

ASRock has started to put out beta BIOSes for their AM4 MBs with the Sinkclose fix.
Edit: MSI MB that was linked, no idea why I wrote ASRock, and separate vulnerability

Ah, no, didn't look close enough, for some previous CVE.

Very similar to the Sinkclose, but via MSI SMM driver that also affects Intel platforms?
jjensn.com/at-home-in-your-firmware/

So all MSI MBs for very long time, AMD and Intel need new BIOS versions to block that.

But not usable from Windows without modifying the OS first.

[visual tree]

Lmao

I've seen people stealing electricity from other neighbours but plugging an electrical device (or a car in this case) directly to a powerline is a whole new level

[languid gulch]

but for about 50 milliseconds, that was the fastest charging tesla on the planet

[visual tree]

I guess you could say the end justifies the means in this case

[soft bloom]

For 20ms assuming they gave 50Hz grid

[visual tree]

Also probably broke the record for the fastest fire ignition on a tesla

[twin dew]

Not really?
As Teslas have had in-car AC to DC chargers for long time.
Just that those depend on external on-off switch for the AC input that is in the wall unit.

So if the pole lines that were tapped illegally were just the normal voltage ones, after the local transformer, then the problem was just that the car couldn't stop charging when full.

Using the DC to AC inverter for the motors as AC to DC charger has been done for long time in various EVs.

[visual tree]

At least the owner wasn't crazy enough to connect a car to transmission lines lol

That probably wouldn't end up well

[twin dew]

Might have been that only after full, or might have been because the AC input was put on before the car was ready...
Forgot there is handshake before the wall box or similar enables the AC too.

Or might have been just that the cars don't have a way to handle the situation where the AC input isn't stopped when they send the signal to cut the power, which would be safety hazard even with normal usage.

The power before the house breakers, but after the last grid stepdown transformer isn't any different from the power after the house breakers.

Just overcurrent protections and the billing measurement unit.

[visual tree]

Considering the articles mentioned houses nearby, It's definitely distribution line with normal voltage (after the last transformer)

[twin dew]

Yup, trying to tap the previous step wouldn't end well.